Aside

logo

Contact

Technical Skills

R STAN Git Docker
Python MATLAB Bash
Markdown Cloud-Computing SQL APIs Binder tidyverse ggplot2 devtools brms caret

Open Source Contributions

I have written and maintain several R packages, available on GitHub:

Main

Granville Matheson

I am an academic data scientist with a background in neuroscience. I am a generalist, but my speciality is in statistical modelling and inference, as well as presentation and communication. My work has made international news and been cited in policy1, I have been involved in developing field-wide guidelines to improve replicability2 and several R packages that I developed are used internationally. I am passionate about learning new things, and enjoy the challenge of presenting complex results in a compelling way to audiences with different backgrounds.

I am currently looking for a position that allows me to work with complex data to derive useful insights, and to develop tools to streamline the process and make it reproducible.

Education

PhD, Neuroscience

Stockholm, Sweden

Karolinska Institutet

2018 - 2014

  • Thesis: Reliability, Replicability and Reproducibility in PET Imaging
  • Working with PET imaging of the dopamine system in psychosis and proneness to developing psychosis, as well as methods development.

MSc, Neuroscience

Utrecht, The Netherlands

Universiteit Utrecht

2013 - 2010

BA Hons, Psychology

Johannesburg, South Africa

University of Witwatersrand

2009

BSc, Psychology, Applied Chemistry

Johannesburg, South Africa

University of Witwatersrand

2008 - 2006

  • Other courses: Chemistry I & II, Major Physics I, Major Pure Mathematics I, Research Design and Analysis

Research Experience

Postdoctoral Researcher*

Columbia University

Molecular Imaging / Biostatistics

2022 - 2020

  • * Cancelled / indefinitely postponed on account of COVID-19 pandemic (NYC)
  • Developing Bayesian methods for performing pharmacokinetic modelling using a multilevel framework, with Markov Chain Monte Carlo.

Postdoctoral Researcher

Karolinska Institutet

Cervenka Lab, PET Group

2020 - 2018

  • Developing tools for reproducible modelling, data storage and documentation.

Research Assistant

Karolinska Institutet

Cervenka Lab, PET Group

2014 - 2012

  • Respondible for image processing and analysis of MR and PET Imaging data for the Karolinska Behavioural PET Database

Research Intern

Universiteit Utrecht

Ramakers Group, Rudolf Magnus Institute

2011 - 2010

  • Worked with single-cell electrophysiology to investigate the dynamics of ion channel caused by morphine

Competencies

  • R package development
  • Bayesian statistics
  • Multilevel modelling
  • Nonlinear models
  • Statistical inference
  • Version control
  • Reproducibility
  • Measurement & Reliability
  • Data visualisation
  • Machine Learning
  • Natural Language Processing
  • Web-scraping
  • API Usage and Deployment

Selected Data Science Writing









I have a blog about data science and visualisation where I publish mostly side projects. Upcoming posts include city commute time visualisations.

Nonlinear Modelling using nls, nlme and brms

granvillematheson.com

N/A

2020

  • A demonstration of how to fit nonlinear models using standard gradient descent optimisation, as well as both frequentist and Bayesian multilevel modelling strategies

Creating an API for 4.2M texts with tidytext, RSQLite, dbplyr and plumber

granvillematheson.com

N/A

2020

  • I prototyped an idea for identification of useful patterns in a large set of text data, to identify useful skills to learn to maximally improve employability from job postings.
  • To handle the data set size, I turned it into a SQLite database for fast, convenient access, and then created an online API to make its results accessible.

Pharmacokinetic Modelling of PET Data in R using kinfitr. Part 2: Basics and Iteration3

granvillematheson.com

N/A

2020

  • Part 2 of a four part series describing my kinetic modelling R package. Here I cover basic usage of the package. I cover bias-variance tradeoffs and other relevant considerations during modelling.

My Physiological Response to my PhD Defence4

granvillematheson.com

N/A

2018

  • I recorded my physiological data in the months leading up to my PhD defence, and analysed it here, using data visualisation to tell the story of my sleep changes, and heart rate, both before and during the defence.
  • I also wrote an R package for extracting this data from the Withings API. I have been contacted by others from around the world who are using my software.

Making a Reminder Bot for Automating Meeting Organisation using R and Google Sheets5

granvillematheson.com

N/A

2018

  • Demonstrated my reminderbot system which has effectively automated the organisation for two sets of meetings continuously for the past 4 years.
  • Setting up a productionised system, running several times each week on a Raspberry Pi

The Weather in Stockholm, Inside and Out, and the Curious Case of Summer 20186

granvillematheson.com

N/A

2018

  • Analyzing meteorological data from open data and private data, examining the effects of global warming, and examining how extreme 2018 really was

Teaching Experience

I am passionate about teaching, and in addition to direct teaching experience, I have held numerous seminars to teach colleagues how to approach statistical problems, and to share knowledge about various new tools that may be helpful.

Positron emission tomography imaging of the CNS

Karolinska Institutet

Stockholm, Sweden

2019 - 2015

  • Lecturer and TA training students in biannual course, teaching about pharmacokinetic modelling and statistical analysis

Research Design and Analysis

University of Witwatersrand

Johannesburg, South Africa

2009

  • Tutor for research design and statistics

Selected Publications

Guidelines for the content and format of PET brain data in publications and archives: A consensus paper

Journal of Cerebral Blood Flow & Metabolism

N/A

2020

  • Authored with all the influential figures in our field
  • This article establishes a consensus for how to report on studies within our field, for which I was asked to contribute my expertise.

Kinfitr - an open source tool for reproducible PET modelling: validation and evaluation of test-retest reliability

bioRxiv

N/A

2020

  • Authored with Jonathan Tjerkaski, Simon Cervenka and Lars Farde
  • I supervised this project, in which we evaluated the performance of my kinetic modelling R package against the established commercial tool used in our field.

The readability of scientific texts is decreasing over time

Elife

N/A

2017

  • Authored with Pontus Plavén-Sigray, Björn Schiffler and William Hedley Thompson
  • This project resulted from our gathering as PhD students, without any supervisors, for a collaborative data science project. In this, we learnt version control with GitHub, collaborative coding among other things.

We need to talk about reliability: making better use of test-retest studies for study design and interpretation

PeerJ

N/A

2015 - 2019

  • Sole author publication
  • I present a new statistical method for estimating study feasibility with limited, and not directly representative data.